CN111208385A - Online fault layered diagnosis method for power grid - Google Patents

Abstract

A power grid online fault layered diagnosis method relates to power grid faults, in particular to a power grid online fault layered diagnosis method which realizes rapid and effective online fault analysis and improves the reliability, economy and safety defense capability of power grid operation. The invention is based on a data acquisition and monitoring control system and relay protection information, comprises four layers of online fault diagnosis and identification technologies of 'switching, SOE event sequence recording, protection action and protection recording', verifies through metering automation, completes identification through a fault knowledge base, and assists a controller to complete analysis of fault types, fault areas, associated equipment and influence loads. The invention establishes a four-layer diagnosis mechanism, realizes quick and effective online fault analysis, is beneficial to striving for time for subsequent fault isolation and fault recovery of scheduling personnel, reduces the loss caused by the power grid fault to the minimum degree, and helps the scheduling personnel to improve the reliability, the economy and the safety defense capability of the power grid operation.

Description

Online fault layered diagnosis method for power grid

Technical Field

The invention relates to a power grid fault, in particular to a power grid online fault layered diagnosis method which can realize rapid and effective online fault analysis and improve the reliability, economy and safety defense capability of power grid operation.

Background

When the power grid fails, a large amount of alarm information is transmitted to each level of power grid dispatching center through the remote terminal devices of all the substations, so that a dispatcher of the regulation and control center can know the real-time power grid operation condition in time. With the continuous enlargement of the scale of the power system and the increasing complexity of the structure, a large amount of alarm information rushes a dispatching center in a short time and far exceeds the processing capacity of operators, so that the dispatching center brings about a lot of troubles for operators on duty, particularly, various types of data such as protection, switching, out-of-limit alarm, abnormal alarm, stable/transient data, fault brief report and the like are heaved up in a fault period, and under the high-voltage environment, mass data can easily show people by terrorism, and misjudgment of dispatchers are easily caused.

Meanwhile, how to reasonably utilize the advantage of the enlargement of the dispatching automation data acquisition scale, the most important information can be quickly and effectively obtained from massive real-time data, high-intelligence, high-speed and high-quality fault analysis service is provided, the silk-spinning and cocoon-peeling of power grid dispatching personnel are assisted, and the comprehensive intelligent analysis and judgment of the power grid fault become an urgent requirement for the regulation and control center to judge and process the fault time period.

Disclosure of Invention

The invention aims to solve the problems that the data volume is large and the wrong judgment and the missed judgment of a dispatcher are easily caused after the existing power grid fails, and provides a power grid online fault layered diagnosis method which can realize the rapid and effective online fault analysis and improve the reliability, the economy and the safety defense capability of the power grid operation.

The invention discloses a layered diagnosis method for online faults of a power grid, which is characterized in that the diagnosis method is based on a data acquisition and monitoring control system and relay protection information, comprises four layers of online fault diagnosis and recognition technologies including 'switch, SOE event sequence recording, protection action and protection recording', carries out verification through metering automation, completes recognition through a fault knowledge base, assists a controller to complete analysis of fault types, fault areas, associated equipment and influence loads, and comprises the following specific analysis steps:

1) first-layer diagnosis, switch-state diagnosis:

when a fault occurs, remote signaling is carried out, namely the action information of a switch is dispatched at first, a fault area is judged according to the tripping condition of a breaker, all permanent faults in the power system are isolated from a normal system through the breaker, and the stable operation of the system is guaranteed;

the relay protection fault parameters comprise a fault phase, a trip phase, a maximum fault current, a maximum zero sequence current and fault distance measurement information, and the fault parameters are actively uploaded depending on action event information or by adopting a general classification data mode and comprise absolute time scale information;

the method comprises the steps that power grid fault diagnosis is carried out by taking a power failure area as a starting condition of fault diagnosis, real-time network topology analysis is carried out under the support of relay protection fault information and a remote signaling displacement signal in a dispatching automation system, a decision tree model of a transformer, a bus, a line and a circuit breaker is established according to main protection, failure protection and backup protection, suspected fault elements are searched according to switch displacement information, and the isolation state of the fault elements in a power system is judged according to a switch state (0,1) value;

based on the above information, a diagnostic model is established:

(1) analyzing and generating a power grid structure based on a power grid CIM model: defining end points and connection points in a CIM (common information model) specification of electric power to express the connection among conductive equipment, buses, disconnecting links, switches and power transmission lines connected among stations, which are connected with each other in a transformer substation, and forming a topological structure of an electric power network;

(2) generating a diagnosis model based on the power grid topological structure: forming a topology incidence matrix of system elements, protection and circuit breakers by combining a power grid operation mode based on a topological structure of a power network, constructing an information fusion diagnosis model, defining a backup protection topology mapping rule and a complete information fusion process according to protection matching logic and an export mode, analyzing the adaptability of the model to network topology change and cascading faults, establishing a main protection, failure protection and backup protection three-layer protection model, and establishing a fault diagnosis model according to the protection model;

(3) judging a fault element according to the displacement switch: after a fault diagnosis model is established, the switch deflection state is monitored in real time on line, and a fault diagnosis program is started after a power failure area is found;

①, the information is discriminated, and various factors are comprehensively considered, so that the accident is prevented from being wrongly identified or misjudged;

② the switch displacement signal, the protection action signal and the power flow disturbance signal have certain relevance, through the switch displacement signal and the topology analysis, after the power failure area is found, the switch displacement property is defined by combining the power flow disturbance information and the protection action information and carrying out comparative analysis, and the switch change property comprises fault disturbance, manual operation and error information;

③, after the switch deflection information is obtained, searching relevant fault models, determining fault characteristics, analyzing the minimum power failure range, and determining a fault element;

2) and a second layer of diagnosis, based on SOE event fault diagnosis:

a joint distribution conditional probability is constructed for a given protection model, protection outlet actions and breaker switching and closing states by utilizing a directed acyclic Bayes network through protection actions and switch displacement information in SOE time, for a given protection model, input and output are represented by nodes, dependence between input and output of basic elements is represented by directed edges between the nodes, the fault occurrence probability P of a basic gate circuit is specified, a conditional probability table and an ideal probability table for each type of basic elements are established to reflect the quantitative relation between input and output of the nodes, and mathematical description is used as follows:

when: x is the number of₁,x₂,…,x_nAre nodes in the same network X, X ═ X₁,x₂,…,x_n) N is a random variable or vector, n events occur simultaneouslyThe probability is:

wherein pi_iDenotes x_iA set of parent nodes;

for each node x in the network_iWhich contains n elementary events (e)₁,e₂,…,e_n) If an event is observed for a node associated therewith in addition to X, the result is E ═ X (X)₁,x₂,…,x_n) Then x_iS event e of_sThe conditional probability of occurrence is:

the switch deflection and the protection action caused by equipment failure are completed within the time T and are sent to a dispatching center monitoring system, then the power grid enters a quiet period, all action signals received within the time range are regarded as signals generated by the failure, the T value is 15 seconds through analyzing and summarizing field failures, all action information generated by the power grid failure is covered within the time period, and the real-time requirement of failure diagnosis is met; reliability evaluation is carried out according to the occurrence time of protection and circuit breaker actions, and the method comprises the following steps:

(1) time reasoning of protection and breaker action time:

defining a time interval T (T)_i)＝[t_i-Δt_i,t_i+Δt_i]Is a time t_iTime range constraint of, Δ t_iThe allowable deviation of the expected action time point of a certain event represents the uncertainty of the occurrence time of the event; definition of d_ij＝t_j-t_iRepresents t_iAnd t_jThe distance in time between the two phases,

time distance range of two time points, Δ d_ijRepresenting uncertainty in length of time;

when event i occurs, event j occurs, if t is known_iAnd

the expected time of occurrence of event j is:

similarly, the occurrence of event i results in the occurrence of event j, if t is known_jAnd

the expected time of occurrence of event i is:

after the actual system fault occurs, the protection and the breaker should act within a setting action time range, and based on the first layer of switch state diagnosis, three layers of protection of the equipment are defined, namely: main, fail and backup protection with corresponding component failure action delay D (t)_c,t_m)、D(t_c,t_p)、D(t_c,t_s) Are respectively [10, 20 ]]、 [600，800]、[1850，2250]Tripping of the circuit breaker with respect to the tripping delay D (t) of the protection outlet_r,t_cb) Is [40, 60 ]]Time delay D (t) of breaker failure protection action relative to action_r,t_f)＝[180，220]In ms;

when reasoning the protection and breaker action time interval, the received first protection outlet action is taken as a reference moment, the time interval of element fault occurrence is obtained by reasoning by combining the protection and breaker action delay, and the expected time interval of the protection and breaker action is obtained by further reasoning;

(2) and (3) calculating the credibility of the action time:

for the protection and circuit breaker action signal, the expected time interval T (t) [ t-delta t, t + delta t ] of the action is obtained by time reasoning]Is a system ofThe action time of the system acquisition is t_i，β_time(i) Representing the credibility of a certain protection and the action time of the breaker; analyzing the expected time interval to accord with normal distribution, wherein the normal distribution is also called Gaussian distribution;

the time of the protection and circuit breaker action signal is brought into a Gaussian function to obtain

When the action time is within the expected time, the reliability is high, otherwise, the reliability is low; two events are involved in the protection and breaker node e₁＝0,e ₁1, when one of the protection or circuit breaker is observed to be inactive, the node e in the model is identified₁Confidence of 0 event mu (e)₁0) equal to state confidence β_state(e₁0); when one of the protection or breaker actions is observed, point e in the model₂Confidence of 1 event mu (e)₂1) includes a state confidence β_state(e₂1) and an operation time reliability β_time(e₂1) with values of:

μ(e₂＝1)＝ω₁β_state(e₂＝1)+ω₂β_time(e₂＝1)

wherein ω is₁、ω₂The weight value representing the state reliability and the action time reliability, the influence of the state of the protection or circuit breaker on the judgment of element faults is larger than that of the time sequence information, and omega is obtained by taking the unit authority to carry out experimental comparison₁、ω₂Values of 0.55 and 0.45;

and increasing the credibility evaluation of the node event occurrence. Mu (e)_i) For the confidence level of the occurrence of an event at a certain node,

x is then_iS event e of_sThe conditional probability of occurrence is:

3) third-level diagnostics, based on PMU protection diagnostics:

representing sinusoidal voltage, current and power of a power system as phasors, voltage

Represents the phasor form:

electric current

Represents the phasor form:

power of

The phasor form represented is:

establishing a reference phasor with the rotation frequency of 50Hz at each station of a power grid by means of a GPS clock signal, and taking other phasors as references to obtain a phase angle;

the sampling pulse of the PMU device is synchronous with a GPS clock, the amplitude and the phase of the input voltage and the current are measured by the full-wave Fourier algorithm to calculate the cycle (20ms), and 25 frames or 50 frames of measured values are provided by the IEEE1344 protocol every second to reflect the dynamic change condition of a power grid; the PMU is also connected with a switching action signal, and when the switching action has no independent time scale, each frame specified by the IEEE1344 protocol has a unified time scale, so that the resolution is 40 ms;

(1) the sampling pulse of the PMU is synchronous with the GPS second pulse, so that the PMU measurement values of different elements are sampled at the same moment when a fault occurs;

(2) the PMU transmits data of one frame every 40ms to a WAMS main station, and the main station detects the change information of element voltage and current from the occurrence of a fault to the removal of the fault;

(3) identifying a fault element by combining the information and the self characteristics of the power system during fault;

the method comprises the following steps of judging a component with a fault by using electric quantity information measured by a PMU through a single-end distance measurement method, carrying out algorithm verification by using a bilateral power supply system equivalent to a power system, and listing a voltage-current relation of an M side when a node M is provided with the PMU according to a wiring diagram:

wherein x is the distance between the fault point and the node M, and Z_lImpedance per unit length of line, R_FIs the excess resistance that is at fault,

is the fault current at point F;

in the formula, the ratio of the total of the components,

for fault currents flowing through the bus M, D_MThe distribution system of the current on the M side is obtained by the following formula:

in the formula, x and R_F、D_AIs an unknown quantity, is multiplied by two equations of an equation

Solving to obtain:

calculating the distance from a fault point to the installation position of the PMU by using the formula according to the electric phasor measured by the PMU to the M section so as to judge whether the detection line has a fault; calculating the measurement distance of the single end of each line within the fault range, determining a branch with a fault, and determining an element of a fault point; and the fault conditions diagnosed by the first layer and the second layer are checked by utilizing the calculation result of the algorithm;

4) and fourth-layer diagnosis, based on the recorded wave data diagnosis:

when the power system has complex faults, further diagnosis is carried out by utilizing fault recording information; the method comprises the steps that the emergency tabulation function of a fault recorder is utilized, the accuracy of fault diagnosis is improved, after a power grid fails, data of each substation fault recorder are transmitted to a dispatching center, data analysis is carried out by operation personnel of the dispatching center to determine a fault element, firstly, the fault element is judged according to the fault diagnosis of the front two layers, wave recording information of a corresponding plant station is checked and transmitted to a central station through a special channel, the central station processes and analyzes the fault property, and if the fault element cannot be determined, the central station of a determined fault area calls the wave recording information of the plant station to analyze, and the fault element and the fault property are determined;

the diagnosis process is based on a multi-agent technology, fault diagnosis based on wave recording data is realized by adopting a distributed cooperation method, a specific diagnosis system comprises a static database, fault wave recording data, system management, data analysis and fault analysis, a power element parameter table, steady-state data and a history are stored in the static database, a configuration file, a fault wave recording real-time data file and static database data of a wave recorder and suspicious elements or areas which are diagnosed by a switch and protection are collected to the system management, fault analysis is carried out after the data analysis, and finally a diagnosis result is output and stored in the history;

and (3) system management: the system has the functions of starting and managing, is responsible for the interaction between the whole system and an external system or a user interface and the reception of data and information, records fault information according to the provided fault information data, queries related fault information by taking a region name, a transformer station name, a wave recorder name and time as key words respectively, and outputs a query result report;

and (3) data analysis: the self-learning function of the multi-agent technology is applied, corresponding data are read from a database according to information provided by a system management agent, a plurality of sub-agent systems of the system are applied, single-end distance measurement, double-end distance measurement, voltage and current waveform analysis, symmetrical component analysis, harmonic analysis and the like are carried out at the same time, and a control agent is responsible for coordination among agents and integration of a final scheme and outputs a diagnosis result;

according to the adaptive function of the agent, the analog quantity and the switching value records of the corresponding wave recorder are extracted, a series of analysis work of the sub-agent systems such as the starting reason of the wave recorder, the correctness of the relay protection action, the disconnection time of the circuit breaker and the like is completed, and the final fault result and the accident reason are given by utilizing the control agent and combining the diagnosis of the first two agent systems.

According to the online fault layered diagnosis method for the power grid, a four-layer diagnosis mechanism is established, so that the online fault analysis is realized quickly and effectively, time is won for subsequent fault isolation and fault recovery of scheduling personnel, loss caused by the power grid fault is reduced to the minimum degree, and the reliability, economy and safety defense capability of the power grid operation are improved.

Drawings

Fig. 1 is a main protection model of a topology.

Fig. 2 is a failure protection model of the topology.

FIG. 3 is a backup fault diagnosis model of a topology.

Fig. 4 is a PMU wiring diagram.

Fig. 5 is a schematic diagram of a diagnostic protocol based on recorded data.

Detailed Description

Example 1: a layered diagnosis method for online faults of a power grid is based on a data acquisition and monitoring control system and relay protection information and comprises four layers of online fault diagnosis and identification technologies including 'switching, SOE event sequence recording, protection action and protection wave recording', verification is carried out through metering automation, identification is completed through a fault knowledge base, and a controller is assisted to complete analysis of fault types, fault areas, associated equipment and influence loads, and the specific analysis steps are as follows:

1) first-layer diagnosis, switch-state diagnosis:

when a fault occurs, remote signaling is carried out, namely the action information of a switch is dispatched at first, a fault area is judged according to the tripping condition of a breaker, all permanent faults in the power system are isolated from a normal system through the breaker, and the stable operation of the system is guaranteed;

the relay protection fault parameters comprise a fault phase, a trip phase, a maximum fault current, a maximum zero sequence current and fault distance measurement information, and the fault parameters are actively uploaded depending on action event information or by adopting a general classification data mode and comprise absolute time scale information;

the method comprises the steps that power grid fault diagnosis is carried out by taking a power failure area as a starting condition of fault diagnosis, real-time network topology analysis is carried out under the support of relay protection fault information and a remote signaling displacement signal in a dispatching automation system, a decision tree model of a transformer, a bus, a line and a circuit breaker is established according to main protection, failure protection and backup protection, suspected fault elements are searched according to switch displacement information, and the isolation state of the fault elements in a power system is judged according to a switch state (0,1) value;

based on the above information, a diagnostic model is established:

(1) analyzing and generating a power grid structure based on a power grid CIM model: defining end points and connection points in a CIM (common information model) specification of electric power to express the connection among conductive equipment, buses, disconnecting links, switches and power transmission lines connected among stations, which are connected with each other in a transformer substation, and forming a topological structure of an electric power network;

(2) generating a diagnosis model based on the power grid topological structure: forming a topology incidence matrix of system elements, protection and circuit breakers by combining a power grid operation mode based on a topological structure of a power network, constructing an information fusion diagnosis model, defining a backup protection topology mapping rule and a complete information fusion process according to protection matching logic and an export mode, analyzing the adaptability of the model to network topology change and cascading faults, establishing a main protection, failure protection and backup protection three-layer protection model, and establishing a fault diagnosis model according to the protection model;

(3) judging a fault element according to the displacement switch: after a fault diagnosis model is established, the switch deflection state is monitored in real time on line, and a fault diagnosis program is started after a power failure area is found;

①, the information is discriminated, and various factors are comprehensively considered, so that the accident is prevented from being wrongly identified or misjudged;

② the switch displacement signal, the protection action signal and the power flow disturbance signal have certain relevance, through the switch displacement signal and the topology analysis, after the power failure area is found, the switch displacement property is defined by combining the power flow disturbance information and the protection action information and carrying out comparative analysis, and the switch change property comprises fault disturbance, manual operation and error information;

③, after the switch deflection information is obtained, searching relevant fault models, determining fault characteristics, analyzing the minimum power failure range, and determining a fault element;

2) and a second layer of diagnosis, based on SOE event fault diagnosis:

a joint distribution conditional probability is constructed for a given protection model, protection outlet actions and breaker switching and closing states by utilizing a directed acyclic Bayes network through protection actions and switch displacement information in SOE time, for a given protection model, input and output are represented by nodes, dependence between input and output of basic elements is represented by directed edges between the nodes, the fault occurrence probability P of a basic gate circuit is specified, a conditional probability table and an ideal probability table for each type of basic elements are established to reflect the quantitative relation between input and output of the nodes, and mathematical description is used as follows:

if x₁,x₂,…,x_nAre nodes in the same network X, X ═ X₁,x₂,…,x_n) And n is a random variable or vector, the probability of the simultaneous occurrence of n events is:

wherein pi_iDenotes x_iA set of parent nodes;

for each node x in the network_iWhich contains n elementary events (e)₁,e₂,…,e_n) If an event is observed for a node associated therewith in addition to X, the result is E ═ X (X)₁,x₂,…,x_n) Then x_iS event e of_sThe conditional probability of occurrence is:

the switch deflection and the protection action caused by equipment failure are completed within the time T and are sent to a dispatching center monitoring system, then the power grid enters a quiet period, all action signals received within the time range are regarded as signals generated by the failure, the T value is 15 seconds through analyzing and summarizing field failures, all action information generated by the power grid failure is covered within the time period, and the real-time requirement of failure diagnosis is met; reliability evaluation is carried out according to the occurrence time of protection and circuit breaker actions, and the method comprises the following steps:

(1) time reasoning of protection and breaker action time:

defining a time interval T (T)_i)＝[t_i-Δt_i,t_i+Δt_i]Is a time t_iTime range constraint of, Δ t_iThe allowable deviation of the expected action time point of a certain event represents the uncertainty of the occurrence time of the event; definition of d_ij＝t_j-t_iRepresents t_iAnd t_jThe distance in time between the two phases,

time distance range of two time points, Δ d_ijRepresenting uncertainty in length of time;

when event i occurs, event j occurs, if t is known_iAnd

the expected time of occurrence of event j is:

similarly, the occurrence of event i results in the occurrence of event j, if t is known_jAnd

the expected time of occurrence of event i is:

after the actual system fault occurs, the protection and the breaker should act within a setting action time range, and based on the first layer of switch state diagnosis, three layers of protection of the equipment are defined, namely: main, fail and backup protection with corresponding component failure action delay D (t)_c,t_m)、D(t_c,t_p)、D(t_c,t_s) Are respectively [10, 20 ]]、 [600，800]、[1850，2250]Tripping of the circuit breaker with respect to the tripping delay D (t) of the protection outlet_r,t_cb) Is [40, 60 ]]Time delay D (t) of breaker failure protection action relative to action_r,t_f)＝[180，220]In ms;

when reasoning the protection and breaker action time interval, the received first protection outlet action is taken as a reference moment, the time interval of element fault occurrence is obtained by reasoning by combining the protection and breaker action delay, and the expected time interval of the protection and breaker action is obtained by further reasoning;

(2) and (3) calculating the credibility of the action time:

for the protection and circuit breaker action signal, the expected time interval T (t) [ t-delta t, t + delta t ] of the action is obtained by time reasoning]The action time of the system acquisition is t_i，β_time(i) Indicating a certain protection and circuit breaker actionReliability of engraving; analyzing the expected time interval to accord with normal distribution, wherein the normal distribution is also called Gaussian distribution;

the time of the protection and circuit breaker action signal is brought into a Gaussian function to obtain

When the action time is within the expected time, the reliability is high, otherwise, the reliability is low; two events are involved in the protection and breaker node e₁＝0,e ₁1, when one of the protection or circuit breaker is observed to be inactive, the node e in the model is identified₁Confidence of 0 event mu (e)₁0) equal to state confidence β_state(e₁0); when one of the protection or breaker actions is observed, point e in the model₂Confidence of 1 event mu (e)₂1) includes a state confidence β_state(e₂1) and an operation time reliability β_time(e₂1) with values of:

μ(e₂＝1)＝ω₁β_state(e₂＝1)+ω₂β_time(e₂＝1)

wherein ω is₁、ω₂The weight value representing the state reliability and the action time reliability, the influence of the state of the protection or circuit breaker on the judgment of element faults is larger than that of the time sequence information, and omega is obtained by taking the unit authority to carry out experimental comparison₁、ω₂Values of 0.55 and 0.45;

and increasing the credibility evaluation of the node event occurrence. Mu (e)_i) For the confidence level of the occurrence of an event at a certain node,

x is then_iS event e of_sThe conditional probability of occurrence is:

3) third-level diagnostics, based on PMU protection diagnostics:

representing sinusoidal voltage, current and power of a power system as phasors, voltage

Represents the phasor form:

electric current

Represents the phasor form:

power of

The phasor form represented is:

establishing a reference phasor with the rotation frequency of 50Hz at each station of a power grid by means of a GPS clock signal, and taking other phasors as references to obtain a phase angle; establishing a reference phasor with the rotation frequency of 50Hz at each station of a power grid by means of a GPS clock signal, and taking other phasors as references to obtain a phase angle;

the sampling pulse of the PMU device is synchronous with a GPS clock, the amplitude and the phase of the input voltage and the current are measured by the full-wave Fourier algorithm to calculate the cycle (20ms), and 25 frames or 50 frames of measured values are provided by the IEEE1344 protocol every second to reflect the dynamic change condition of a power grid; the PMU is also connected with a switching action signal, and when the switching action has no independent time scale, each frame specified by the IEEE1344 protocol has a unified time scale, so that the resolution is 40 ms;

(1) the sampling pulse of the PMU is synchronous with the GPS second pulse, so that the PMU measurement values of different elements are sampled at the same moment when a fault occurs;

(2) the PMU transmits data of one frame every 40ms to a WAMS main station, and the main station detects the change information of element voltage and current from the occurrence of a fault to the removal of the fault;

(3) identifying a fault element by combining the information and the self characteristics of the power system during fault;

the method comprises the following steps of judging a component with a fault by using electric quantity information measured by a PMU through a single-end distance measurement method, carrying out algorithm verification by using a bilateral power supply system equivalent to a power system, and listing a voltage-current relation of an M side when a node M is provided with the PMU according to a wiring diagram:

wherein x is the distance between the fault point and the node M, and Z_lImpedance per unit length of line, R_FIs the excess resistance that is at fault,

is the fault current at point F;

in the formula, the ratio of the total of the components,

for fault currents flowing through the bus M, D_MThe distribution system of the current on the M side is obtained by the following formula:

in the formula, x and R_F、D_AIs an unknown quantity, is multiplied by two equations of an equation

Solving to obtain:

calculating the distance from a fault point to the installation position of the PMU by using the formula according to the electric phasor measured by the PMU to the M section so as to judge whether the detection line has a fault; calculating the measurement distance of the single end of each line within the fault range, determining a branch with a fault, and determining an element of a fault point; and the fault conditions diagnosed by the first layer and the second layer are checked by utilizing the calculation result of the algorithm;

4) and fourth-layer diagnosis, based on the recorded wave data diagnosis:

when the power system has complex faults, further diagnosis is carried out by utilizing fault recording information; the method comprises the steps that the emergency tabulation function of a fault recorder is utilized, the accuracy of fault diagnosis is improved, after a power grid fails, data of each substation fault recorder are transmitted to a dispatching center, data analysis is carried out by operation personnel of the dispatching center to determine a fault element, firstly, the fault element is judged according to the fault diagnosis of the front two layers, wave recording information of a corresponding plant station is checked and transmitted to a central station through a special channel, the central station processes and analyzes the fault property, and if the fault element cannot be determined, the central station of a determined fault area calls the wave recording information of the plant station to analyze, and the fault element and the fault property are determined;

the diagnosis process is based on a multi-agent technology, fault diagnosis based on wave recording data is realized by adopting a distributed cooperation method, a specific diagnosis system comprises a static database, fault wave recording data, system management, data analysis and fault analysis, a power element parameter table, steady-state data and a history are stored in the static database, a configuration file, a fault wave recording real-time data file and static database data of a wave recorder and suspicious elements or areas which are diagnosed by a switch and protection are collected to the system management, fault analysis is carried out after the data analysis, and finally a diagnosis result is output and stored in the history;

and (3) system management: the system has the functions of starting and managing, is responsible for the interaction between the whole system and an external system or a user interface and the reception of data and information, records fault information according to the provided fault information data, queries related fault information by taking a region name, a transformer station name, a wave recorder name and time as key words respectively, and outputs a query result report;

and (3) data analysis: the self-learning function of the multi-agent technology is applied, corresponding data are read from a database according to information provided by a system management agent, a plurality of sub-agent systems of the system are applied, single-end distance measurement, double-end distance measurement, voltage and current waveform analysis, symmetrical component analysis, harmonic analysis and the like are carried out at the same time, and a control agent is responsible for coordination among agents and integration of a final scheme and outputs a diagnosis result;

according to the adaptive function of the agent, the analog quantity and the switching value records of the corresponding wave recorder are extracted, a series of analysis work of the sub-agent systems such as the starting reason of the wave recorder, the correctness of the relay protection action, the disconnection time of the circuit breaker and the like is completed, and the final fault result and the accident reason are given by utilizing the control agent and combining the diagnosis of the first two agent systems.

Claims (1)

1. A layered diagnosis method for online faults of a power grid is characterized in that the diagnosis method is based on a data acquisition and monitoring control system and relay protection information, comprises four layers of online fault diagnosis and recognition technologies including 'switching, SOE event sequence recording, protection action and protection recording', carries out verification through metering automation, completes recognition through a fault knowledge base, assists a controller to complete analysis of fault types, fault areas, associated equipment and influence loads, and specifically comprises the following analysis steps:

1) first-layer diagnosis, switch-state diagnosis:

when a fault occurs, remote signaling is carried out, namely the action information of a switch is dispatched at first, a fault area is judged according to the tripping condition of a breaker, all permanent faults in the power system are isolated from a normal system through the breaker, and the stable operation of the system is guaranteed;

the relay protection fault parameters comprise a fault phase, a trip phase, a maximum fault current, a maximum zero sequence current and fault distance measurement information, and the fault parameters are actively uploaded depending on action event information or by adopting a general classification data mode and comprise absolute time scale information;

the method comprises the steps that power grid fault diagnosis is carried out by taking a power failure area as a starting condition of fault diagnosis, real-time network topology analysis is carried out under the support of relay protection fault information and a remote signaling displacement signal in a dispatching automation system, a decision tree model of a transformer, a bus, a line and a circuit breaker is established according to main protection, failure protection and backup protection, suspected fault elements are searched according to switch displacement information, and the isolation state of the fault elements in a power system is judged according to a switch state (0,1) value;

based on the above information, a diagnostic model is established:

(1) analyzing and generating a power grid structure based on a power grid CIM model: defining end points and connection points in a CIM (common information model) specification of electric power to express the connection among conductive equipment, buses, disconnecting links, switches and power transmission lines connected among stations, which are connected with each other in a transformer substation, and forming a topological structure of an electric power network;

(2) generating a diagnosis model based on the power grid topological structure: forming a topology incidence matrix of system elements, protection and circuit breakers by combining a power grid operation mode based on a topological structure of a power network, constructing an information fusion diagnosis model, defining a backup protection topology mapping rule and a complete information fusion process according to protection matching logic and an export mode, analyzing the adaptability of the model to network topology change and cascading faults, establishing a main protection, failure protection and backup protection three-layer protection model, and establishing a fault diagnosis model according to the protection model;

(3) judging a fault element according to the displacement switch: after a fault diagnosis model is established, the switch deflection state is monitored in real time on line, and a fault diagnosis program is started after a power failure area is found;

①, the information is discriminated, and various factors are comprehensively considered, so that the accident is prevented from being wrongly identified or misjudged;

② the switch displacement signal, the protection action signal and the power flow disturbance signal have certain relevance, through the switch displacement signal and the topology analysis, after the power failure area is found, the switch displacement property is defined by combining the power flow disturbance information and the protection action information and carrying out comparative analysis, and the switch change property comprises fault disturbance, manual operation and error information;

③, after the switch deflection information is obtained, searching relevant fault models, determining fault characteristics, analyzing the minimum power failure range, and determining a fault element;

2) and a second layer of diagnosis, based on SOE event fault diagnosis:

a joint distribution conditional probability is constructed for a given protection model, protection outlet actions and breaker switching and closing states by utilizing a directed acyclic Bayes network through protection actions and switch displacement information in SOE time, for a given protection model, input and output are represented by nodes, dependence between input and output of basic elements is represented by directed edges between the nodes, the fault occurrence probability P of a basic gate circuit is specified, a conditional probability table and an ideal probability table for each type of basic elements are established to reflect the quantitative relation between input and output of the nodes, and mathematical description is used as follows:

if x₁，x₂，...，x_nAre nodes in the same network X, X ═ X₁，x₂，...，x_n) And n is a random variable or vector, the probability of the simultaneous occurrence of n events is:

wherein pi_iDenotes x_iA set of parent nodes;

for each node x in the network_iWhich contains n elementary events (e)₁，e₂，...，e_n) If an event is observed for a node associated therewith in addition to X, the result is E ═ X (X)₁，x₂，...，x_n) Then x_iS event e of_sThe conditional probability of occurrence is:

the switch deflection and the protection action caused by equipment failure are completed within the time T and are sent to a dispatching center monitoring system, then the power grid enters a quiet period, all action signals received within the time range are regarded as signals generated by the failure, the T value is 15 seconds through analyzing and summarizing field failures, all action information generated by the power grid failure is covered within the time period, and the real-time requirement of failure diagnosis is met; reliability evaluation is carried out according to the occurrence time of protection and circuit breaker actions, and the method comprises the following steps:

(1) time reasoning of protection and breaker action time:

defining a time interval T (T)_i)＝[t_i-Δt_i，t_i+Δt_i]Is a time t_iTime range constraint of, Δ t_iThe allowable deviation of the expected action time point of a certain event represents the uncertainty of the occurrence time of the event; definition of d_ij＝t_j-t_iRepresents t_iAnd t_jThe distance in time between the two phases,

time distance range of two time points, Δ d_ijRepresenting uncertainty in length of time;

when event i occurs, event j occurs, if t is known_iAnd

the expected time of occurrence of event j is:

similarly, the occurrence of event i results in the occurrence of event j, if t is known_jAnd

the expected time of occurrence of event i is:

after the actual system fault occurs, the protection and the breaker should act within a setting action time range, and based on the first layer of switch state diagnosis, three layers of protection of the equipment are defined, namely: main, fail and backup protection with corresponding component failure action delay D (t)_c，t_m)、D(t_c，t_p)、D(t_c，t_s) Are respectively [10, 20 ]]、[600，800]、[1850，2250]Tripping of the circuit breaker with respect to the tripping delay D (t) of the protection outlet_r，t_cb) Is [40, 60 ]]Time delay D (t) of breaker failure protection action relative to action_r，t_f)＝[180，220]In ms;

when reasoning the protection and breaker action time interval, the received first protection outlet action is taken as a reference moment, the time interval of element fault occurrence is obtained by reasoning by combining the protection and breaker action delay, and the expected time interval of the protection and breaker action is obtained by further reasoning;

(2) and (3) calculating the credibility of the action time:

for the protection and circuit breaker action signal, the expected time interval T (t) [ t-delta t, t + delta t ] of the action is obtained by time reasoning]The action time of the system acquisition is t_i，β_time(i) Representing the credibility of a certain protection and the action time of the breaker; analyzing the expected time interval to accord with normal distribution, wherein the normal distribution is also called Gaussian distribution;

the time of the protection and circuit breaker action signal is brought into a Gaussian function to obtain

When the action time is within the expected time, the reliability is high, otherwise, the reliability is low; two events are involved in the protection and breaker node e₁＝0，e₁1, when one of the protection or circuit breaker is observed to be inactive, the node e in the model is identified₁Confidence of 0 event mu (e)₁0) equal to state confidence β_state(e₁0); when one of the protection or breaker actions is observed, point e in the model₂Confidence of 1 event mu (e)₂1) includes a state confidence β_state(e₂1) and an operation time reliability β_time(e₂1) with values of:

μ(e₂＝1)＝ω₁β_state(e₂＝1)+ω₂β_time(e₂＝1)

wherein ω is₁、ω₂The weight value representing the state reliability and the action time reliability, the influence of the state of the protection or circuit breaker on the judgment of element faults is larger than that of the time sequence information, and omega is obtained by taking the unit authority to carry out experimental comparison₁、ω₂Values of 0.55 and 0.45;

and increasing the credibility evaluation of the node event occurrence. Mu (e)_i) For the confidence level of the occurrence of an event at a certain node,

x is then_iS event e of_sThe conditional probability of occurrence is:

3) third-level diagnostics, based on PMU protection diagnostics:

representing sinusoidal voltage, current and power of a power system as phasors, voltage

Represents the phasor form:

electric current

Represents the phasor form:

power of

The phasor form represented is:

establishing a reference phasor with the rotation frequency of 50Hz at each station of a power grid by means of a GPS clock signal, and taking other phasors as references to obtain a phase angle;

the sampling pulse of the PMU device is synchronous with a GPS clock, the amplitude and the phase of the input voltage and the current are measured by the full-wave Fourier algorithm to calculate the cycle (20ms), and 25 frames or 50 frames of measured values are provided by the IEEE1344 protocol every second to reflect the dynamic change condition of a power grid; the PMU is also connected with a switching action signal, and when the switching action has no independent time scale, each frame specified by the IEEE1344 protocol has a unified time scale, so that the resolution is 40 ms;

(1) the sampling pulse of the PMU is synchronous with the GPS second pulse, so that the PMU measurement values of different elements are sampled at the same moment when a fault occurs;

(2) the PMU transmits data of one frame every 40ms to a WAMS main station, and the main station detects the change information of element voltage and current from the occurrence of a fault to the removal of the fault;

(3) identifying a fault element by combining the information and the self characteristics of the power system during fault;

the method comprises the following steps of judging a component with a fault by using electric quantity information measured by a PMU through a single-end distance measurement method, carrying out algorithm verification by using a bilateral power supply system equivalent to a power system, and listing a voltage-current relation of an M side when a node M is provided with the PMU according to a wiring diagram:

wherein x is the distance between the fault point and the node M, and z_lImpedance per unit length of line, R_FIs the excess resistance that is at fault,

is the fault current at point F;

in the formula, the ratio of the total of the components,

for fault currents flowing through the bus M, D_MThe distribution system of the current on the M side is obtained by the following formula:

in the formula, x and R_F、D_AIs an unknown quantity, is multiplied by two equations of an equation

Solving to obtain:

calculating the distance from a fault point to the installation position of the PMU by using the formula according to the electric phasor measured by the PMU to the M section so as to judge whether the detection line has a fault; calculating the measurement distance of the single end of each line within the fault range, determining a branch with a fault, and determining an element of a fault point; and the fault conditions diagnosed by the first layer and the second layer are checked by utilizing the calculation result of the algorithm;

4) and fourth-layer diagnosis, based on the recorded wave data diagnosis:

when the power system has complex faults, further diagnosis is carried out by utilizing fault recording information; the method comprises the steps that the emergency tabulation function of a fault recorder is utilized, the accuracy of fault diagnosis is improved, after a power grid fails, data of each substation fault recorder are transmitted to a dispatching center, data analysis is carried out by operation personnel of the dispatching center to determine a fault element, firstly, the fault element is judged according to the fault diagnosis of the front two layers, wave recording information of a corresponding plant station is checked and transmitted to a central station through a special channel, the central station processes and analyzes the fault property, and if the fault element cannot be determined, the central station of a determined fault area calls the wave recording information of the plant station to analyze, and the fault element and the fault property are determined;

the diagnosis process is based on a multi-agent technology, fault diagnosis based on wave recording data is realized by adopting a distributed cooperation method, a specific diagnosis system comprises a static database, fault wave recording data, system management, data analysis and fault analysis, a power element parameter table, steady-state data and a history are stored in the static database, a configuration file, a fault wave recording real-time data file and static database data of a wave recorder and suspicious elements or areas which are diagnosed by a switch and protection are collected to the system management, fault analysis is carried out after the data analysis, and finally a diagnosis result is output and stored in the history;

and (3) system management: the system has the functions of starting and managing, is responsible for the interaction between the whole system and an external system or a user interface and the reception of data and information, records fault information according to the provided fault information data, queries related fault information by taking a region name, a transformer station name, a wave recorder name and time as key words respectively, and outputs a query result report;

and (3) data analysis: the self-learning function of the multi-agent technology is applied, corresponding data are read from a database according to information provided by a system management agent, a plurality of sub-agent systems of the system are applied, single-end distance measurement, double-end distance measurement, voltage and current waveform analysis, symmetrical component analysis, harmonic analysis and the like are carried out at the same time, and a control agent is responsible for coordination among agents and integration of a final scheme and outputs a diagnosis result;

according to the adaptive function of the agent, the analog quantity and the switching value records of the corresponding wave recorder are extracted, a series of analysis work of the sub-agent systems such as the starting reason of the wave recorder, the correctness of the relay protection action, the disconnection time of the circuit breaker and the like is completed, and the final fault result and the accident reason are given by utilizing the control agent and combining the diagnosis of the first two agent systems.

Images